Study On Surrogate Assisted CMA-ES And Its Application In Composite Material Design

Composite materials are a kind of material with high specific strength,high specific stiffness,high specific modulus and material properties.Therefore,composite materials are becoming more and more widely used in the automotive industry.As one of the basic components of automobile structure,the research of composite laminates is also getting more and more researchers.Pay attention.Evolutionary algorithm is a universal,stable,and globally convergent optimization method.However,when optimizing the actual engineering problems,a large number of positive problems need to be calculated,resulting in the optim ization process taking a long time and low efficiency.By comparing a number of evolutionary algorithms,this study considers the Covariance matrix adaptation evolution strategies(CMA-ES)and the agent model to reduce the cost of computing.Test functions and examples also show that the CMA-ES evolutionary algorithm based on surrogate model is much better than other similar algorithms in terms of computational efficiency.The specific work of this article includes:1)CMA-ES evolutionary algorithm based on Proxy ModelCMA-ES algorithm and Least squares support vector machine regression(LS-SVR)are compiled.Comparing the CMA-ES evolutionary algorithm and the LS-SVR proxy model with the same type of algorithm,the superiority of the algorithm and the model is ensured.Considering the dependence of the global model on the number of samples,a strategy based on local dynamic modeling is proposed,that is,the established agent model only needs to ensure that the search efficiency of the current iterative proce ss is accurate enough,and then the search efficiency of CMA-ES can be greatly improved.2)LSS-CMA-ES test and verificationUsing 13 groups of test functions that can represent all kinds of problems,compare the LSS-CMA-ES with the same type of Surrogate assisted evolution algorithms(SAEAS),verify the superiority of LSS-CMA-ES in the same type algorithm,and then compare the LSS-CMA-ES with the mainstream CMA-ES,and verify that the LSS-CMA-ES is in the same type.The importance of the agent model in this algorithm.3)research on the performance evaluation standard of the algorithmComparing the accuracy and algorithm performance of the LS-SVR proxy model,we get the conclusion that the performance of the algorithm is positively related to the accuracy of the proxy model in the algorithm.Then the performance of the algorithm is judged by observing the distribution trend of the scores related to the accuracy of the agent model in the iterative process.4)global optimization of composites based on Isogeo metric analysis(IGA)using LSS-CMA-ES.The IGA solver based on composite design is compiled,and the seamless connection between the optimization algorithm and the solver is realized.LSS-CMA-ES and CMA-ES are used to optimize the structure of the following problems,including the design goal of the maximum self vibration frequency of the laminated plate,and the design of the layer angle of the composite laminates.The design of the composite laminates is designed by the minimization of the maximum stress of the laminates as the design goal.Taking the maximum stress minimization of tubular structure as the design objective,the scaling factor of radius is taken as the design variable to optimize the design.